Fault diagnosis of bearings based on an improved lightweight convolution neural network

Great progresses have been made in fault diagnosis of bearings based on convolutional neural networks, but these models bring a significant burden on the hardware, increase industrial costs, and inconvenience to the updating and training of models. A good fault diagnosis model should have a low number of parameters and be able to achieve high accuracy. In order to better reduce the number of network parameters while maintaining high accuracy, this study proposes a lightweight network model that can solve both of these problems through experimental comparison.

Automated summary

In this study, a lightweight network model is proposed to address the issues of low parameter count and high accuracy through experimental comparison.